Economics Program Working Paper Series

1 Economics Program 845 Third Avenue New York, NY 10022-6600 Tel. 212-339-0420 Economics Program Working Paper Series Comprehensive Benchmark Revisions for The Conference Board Leading economic Index for the United States Gad Levanon Ataman Ozyildirim Brian Schaitkin Justyna Zabinska The Conference Board December 2011 EPWP #11 06 1 Comprehensive Benchmark Revisions of The Conference Board Leading economic Index for the United States1 By Gad Levanon Ataman Ozyildirim Brian Schaitkin Justyna Zabinska The Conference Board December 2011 Abstract Following an extensive reevaluation of existing indicators included in The Conference Board Leading economic Index for The United States, we propose a comprehensive revision of the composite index. In this Paper we present the case for replacing three of the components and making a minor adjustment to one other component. The resulting index addresses structural changes that have occurred in the economy in the last several decades.

2 The changes in the LEI composition include: 1) incorporating in the LEI a new Leading Credit Index (LCI) rather than real money supply (M2) starting in 1990 (real M2 remains in the index before 1990); 2) replacing the ISM Supplier Delivery Index with the ISM New Orders Index; 3) replacing the Reuters/University of Michigan Consumer Expectations Index with an equally weighted average of consumer expectations of business and economic conditions using questions from surveys conducted by Reuters/University of Michigan and The Conference Board; and 4) replacing New Orders for (nondefense) Capital Goods with New Orders for (nondefense) Capital Goods excluding Aircraft. These changes are assessed using turning point analysis, probit models and an indicator scoring system based on Markov Switching models. Real time out-of-sample forecasting exercises are used to confirm that the changes to the composition help the LEI forecast more accurately future economic conditions.

3 Keywords: business cycles, turning points, leading economic indexes, Markov Switching, probit, out-of-sample forecasting, diffusion indexes 1 The Conference Board, Inc. 2011. Corresponding author: Ataman Ozyildirim, We would like to thank members of The Conference Board Business Cycle Indicators Advisory Panel,for helpful comments and suggestions. We would also like to thank Jennelyn Tanchua for excellent research assistance on an earlier version. All remaining errors are, of course, ours. The views expressed in this Paper are those of the author(s) and do not necessarily represent those of The Conference Board. 2 I. Introduction The last comprehensive revision of the leading economic index (LEI) for the United States was implemented in 1996, after TCB had assumed responsibility for the Business Cycle Indicators Program and started publishing the LEI. The 1996 revision introduced the interest rate spread as a measure which signals recessions through information form Treasury markets and the stance of monetary policy (the LEI component is the difference between the 10-year Treasury bond yields and the Federal Funds rate, set by the Federal Reserve).

4 Other 1996 revisions involved changing or adjusting the measures used to cover manufacturing orders, commodity prices, and inflation. Since then, some methodological changes have been implemented in 2001 and 2005 as well, such as the re-introduction of trend adjustment, and the implementation of a new calculation method for the contribution of the yield spread component. (These revisions are documented in benchmark articles on the TCB web site.) In March 2010 The Conference Board published an article, titled Real M2 and Its Impact on The Conference Board Leading economic Index (LEI) for the United States in which it communicated that it was considering removing real M2 from the LEI and replacing it with an indicator of financial conditions. Further research on this subject, the results of which and final recommendations for changes in the composition of the LEI, is discussed separately in a companion Working Paper entitled Using a Composite Index of Financial Conditions Indicators to Predict Turning Points in the Business Cycle, by Levanon et.

5 Al. (2011) posted on The Conference Board web site. The major recommendation of this research is that because the real money supply component (real M2) has ceased to perform well as a leading indicator it should be omitted and that a newly developed Leading Credit Index (LCI) be incorporated into the LEI. This Paper presents the results of our review of the other components of the LEI. It also compares the performance of an alternative index including the LCI and the other recommended revisions compared with the current index. 3 II. Revisions to the Composition of the LEI The LEI is currently made up of ten components. Leaving the real money supply component aside (which is discussed in the Working Paper mentioned above), our recommendations affect the following leading indicators (1) Manufacturing New Orders for (non-defense) Capital Goods, (2) ISM Index of Supplier Deliveries and, (3) Consumer Expectations. Below are the changes, which will be discussed in greater detail in the following sections2.

6 Replace New Orders for (nondefense) Capital Goods with New Orders for (nondefense) Capital Goods excluding Aircraft . Replace ISM Supplier Deliveries Index ( vendor performance index) with the ISM New Orders Index for Manufacturing. In addition, it will be the level of this component, rather than its change, that contributes to the LEI (similar to the current approach for the interest rate spread and the new leading credit index components). Replace the Reuters/University of Michigan Consumer Expectations Index with a new component - a combination of Consumer expectations of Business and economic conditions from the surveys conducted by Reuters/University of Michigan and The Conference Board. It will also be the level of this component that contributes to the index. Methodology To analyze the current components and select potential substitutes, we employed an approach that is based on probability models (probit models3 and/or models based on Markov Switching4) 2 Van Dijk(2011) determines using a Bayesian estimation procedure allowing the LEI components to be weighted unequally that both the ISM Index of Supplier Deliveries and M2 would receive almost no weight on average.

7 3 The capacity of leading indicators to anticipate recessions can be tested by incorporating them into probit models. The structure of these models is to use the proposed leading indicators as t-quarters ahead lagged independent variables, where a binary dependent variable takes the value of 1 when economy is in recession and 0 when it is in expansion. This procedure will generate a Series of predictions based upon the behavior of the variable describing the likelihood of a recession in the quarter t-periods into the future. The quality of the forecasts is measured by calculating an error term as the difference between the predicted recession probability and the binary value stating whether a recession occurred during that quarter. From these error values, a 4 to evaluate and score an indicator's ability to forecast turning points. We, then, ranked the different indicators based on these results. We also used turning point analysis used in The Conference Board indicator approach to supplement these methodologies.

8 However, the turning point analysis was not very applicable to several indicators we looked at, since it is the levels of these Series that provide information on the cyclical outlook. These indicators (often based on diffusion levels) would have to be cumulated around a threshold to make the indicator comparable to business cycle measure in levels. To confirm that the changes in composition produces a better index, the current LEI and an alternative LEI are compared in terms of forecasting ability using real time out-of-sample forecasting exercises (for a discussion and empirical results on the effects of composition changes in the LEI see McGuckin and Ozyildirim, 2004). quadratic probability score (QPS) can be calculated and compared across different variables and their recession forecasts.

9 4 Markov Switching models are based on the idea that the parameters of an econometric model are not constant over time and should instead be allowed to shift between multiple states. At each observation, the probability that a given variable is in the low or high regime state can be measured. Leading indicators should move into the low regime state in advance of a business cycle peak and remain there until the trough of the business cycle is approaching. In our approach, the way this method is used for evaluating leading indicators compares the timing of the periods with the highest low-regime probabilities with the timing of recessions. For example, in the 1959-2011 period, there were 34 quarters that are considered recessions. During that time, we compare the timing of the 34 quarters with the highest low-regime probabilities for each indicator with the timing of the recession quarters. We choose the same number of recession signal quarters, 34, as the number of quarters in recessions, because if we demand that leading indicators signal both peaks and troughs, then the duration of the recession signal needs to start before the peak and end before the trough.

10 That means that the duration of the recession signal is roughly the same as the recession itself. We divide the sample into good zones and bad zones. The good zone is a period where we would want a good leading indicator to signal a recession. In this method we defined the good zone as the zone that includes the three quarters prior to the beginning of the recession and quarters during the recession except for the last two quarters of the recession. The bad zone is a period between the last quarter of a recession and four quarters prior to the next recession. One quarter before the last quarter of the recession is a neutral zone because it is not clear if a good leading indicator should signal a recession during that quarter. For a more detailed description of the method, please see Levanon (2010). 5 Performance of the LEI Components Table 1 shows the ranking of the quadratic probability scores (QPS), calculated for the LEI and its ten current components.